📄 Showing the "Simple" version — "Popular" is not ready yet. Add it to favorites to help prioritize it. Ask for the full read

Viscoelastic materials — polymers that, like dough, retain a memory of every touch: their response to impact depends on their deformation history. Maxwell and Boltzmann provided the mathematical description long ago. But for computer simulations, engineers had to model each molecule at every step — an impossible task even for supercomputers.

Now a neural network trained on thousands of virtual experiments comes to the rescue. It captures patterns without iterating over billions of particles, much like a skilled baker who, just by looking at dough, understands its entire history. The network's internal parameters accumulate information about the past, much like entropy gathers data on molecular disorder.

Tests show the model accurately reproduces the behavior of a carbon-based polymer, polyurea, including under cyclic bending, impact, and even considering heating. A remarkable fact: a layer of this polyurea just a few millimeters thick can shield a concrete wall from a blast wave. The neural net works thousands of times faster than classical models and can be easily integrated into any engineering software as a universal template — analogous to the standard model for materials physics.